- Core Courses: Deep Learning, Time Series Analysis, Digital Image Processing, C++, Modern Control Theory.
- Core Courses: Finance Engineering, Statistics, Accounting, Macroeconomics, Management.
- Honors: National Brandstorm Top 10, Brandstorm Second prize in New York University, Shanghai
- Used Python to segment text by Jieba to quantitatively analyze the frequency on major events and slogans in various provinces.
- Frequency of keywords in text proved that historical wheat-farming regions emphasize independent goals for adolescent development, while rice-farming regions emphasize interdependence.
- Implemented the code from Deep ADMM-Net and added Structural Similarity Index as a measurement in the training process.
- Visualized the PSNR-epoch curve and image iteration process through the Deep ADMM-Net.
- Used Python to write a web scrawler to gather flight, weather information in Flightware.
- Recode Random forest from the bottom layer, with the max depth of five. The result of random forest classification in forecasting whether your flight would delay or not reached over 95%.
- Gathered classic poetries from prestigious poets and a data pool of over 400 classic images and their classic meanings from Tang Dynasty to 1970 as our training data and gathered modern poetries over 40,000 words from 1970 untill now as test data.
- Predicted meanings of imageries in test data with an accuracy of over 80%.
- Led a nickname extraction test project by Natural Language Processing, aiming at collecting actors’ or roles’ nicknames through bullets in TV shows or movies, which reached an accuracy of 95% and a recall rate of 89.9% by Java, Python, C, and Shell.
- Analyzed the user profiling on the aspects of aging and gender by Python and Hadoop, and categorized over 600 TV shows from 2017, as a feature in forecasting model of play amount, which reached an accuracy over 90%.
- Developed a chatbot by Java and Hadoop, using Natural Language Processing techniques to translate natural language toMySQL searching query and applying Elastic Search to enhance the results.
- Forecasted the TV show after the first, second, third, seventh day of its launch using GBDT, whose MAPE loss maintains within 20%. Optimized the algorithm and the process of data processing by over 50% using Hive, Java, Babel and Shell.
- Wrote SQL to gather data from departments and automatized the monthly report to twenty-two company brands by Python.
- Forecasted the sale of double eleven for Lancome with Time-Series data over two years and reached precision of 95% by Xgboost and Machine Learning, which saved time for company to ship goods and boosted sales in double eleven by 20%.
- Held lectures of Big data about databank, a data analysis website designed by our team with over one hundred people present.
- Translated over twenty paper and news of top-notch technologies into Chinese, including the State of Deep Learning in 2018, and responsible for the weekly AI Scholar module. Accumulated reading quantity reached over one hundred thousand.
- Led college publicity activities, including activity poster design and college social media platform operation.
- Organized a lecture of Super brain, with over two hundred student participants.
- Interviewed professionals in Machine Learning about their own successful experience, their opinion of the future for Machine Learning, and their own notion of building up a geek organization with leading-edge workshops.
Ranked 10th in China
- Used Baidu API to recognize food in the table, just by photographing your meal, and immediately give you the calories of the food.
- Conforming to the topic of competition, Intelligent Furnitures, with the help of intelligent fridge, we can recommend food for users and post food ingredients to users on the early morning based on their eating habits.
Top 10 in China, and ranked 2rd in Shanghai
- Personalize your salon experience by inventing a kit, like 23&me, to detect physical property, chemical property of your hair situation, and sending back samples to promote products’ sales.
- Created an app to shop this kit online and read your personal report within 2 weeks.
- Rescued citizens in Puerto Rico by solving Travelling Salesman Problem by Markov Chain using Java.
- Found the perfect base for the rescue team by KNN algorithm and packed needed medicines into a 3D package by 3D Greedy Algorithm using Python.
- Using logistic regression to forecast the population of citizens who use ten languages separately in 15 years.
- Using CNN to forecast the population of citizens who use ten languages separately in 50 years.
- Solved Travelling Salesman Problem, a classic NP hard problem by Minimum Spanning Tree Algorithm.
- Visualized the result of routes within the city by Web.
- Computer: Proficient in Python, MATLAB, Java, MySQL, Shell(Linux), LaTeX, EXCEL, understand C++, Hadoop
- Interests: Amateur landscape photographer, Music, Networking.